JAY Project
JAY Project

How Accurate Is a Guitar Tuner? — 440Hz, Cents, and Why Low Bass Notes Read an Octave High

2026-09-14| Jay

The needle centers, the light turns green, and you assume you're in tune. But how in tune does that green light actually mean?

This post walks through what a tuner measures, how fine those numbers are, and why low notes so often read an octave too high. (That last one turned up in this site's own tuner while writing, and was fixed the same day.)

1. Everything Comes From One Number: A4 = 440Hz

Tuning has exactly one anchor: A4 vibrates 440 times per second. Every other note is calculated from it.

The rule is simple. Each semitone multiplies the frequency by the 12th root of 2.

2^(1/12) = 1.05946...

A4  = 440.00 Hz
A#4 = 440 × 1.05946 = 466.16 Hz
B4  = 440 × 1.05946² = 493.88 Hz
...
A5  = 440 × 1.05946¹² = 880 Hz  (exactly double)

Twelve multiplications land exactly on double. One octave = double the frequency is where the whole system starts.

That's why guitar strings sit on odd-looking numbers — low E2 at 82.41Hz, high E4 at 329.63Hz. You don't need to memorize them; just know they're derived from 440.

2. Cents — A Semitone Cut Into 100 Pieces

When a tuner shows "+3" or "−7", the unit is cents.

1 semitone = 100 cents
1 octave   = 1200 cents

Converting a frequency ratio into cents:

cents = 1200 × log₂(measured ÷ target)

The logarithm is there because pitch moves by multiplication, not addition. Raising 440Hz by 1Hz and raising 880Hz by 1Hz sound completely different — about 3.9 cents versus about 2 cents.

Cents erase that difference so one ruler works across every register.

3. So How Many Cents Is "Green"?

The tuner on this site turns green within ±5 cents. The meter runs out at ±50 cents. Since 50 cents is half a semitone, a pegged needle means you're roughly halfway to the neighbouring note.

For a sense of scale:

Comparison Difference
440Hz vs 441Hz about 3.9 cents
440Hz vs 442Hz about 7.9 cents
One semitone 100 cents

Human discrimination between two pitches heard back-to-back sits around 5–10 cents. So ±5 cents is calibrated to roughly where the ear stops being able to argue.

One practical consequence follows. Some ensembles tune to 442Hz instead of 440, and that 2Hz gap is about 7.9 cents — past the green threshold. Show up to a 442 rehearsal tuned to 440 and you'll be correct by the tuner but wrong in the room.

That's why tuners let you move the reference pitch. On this tool, shifting the reference moves every string's target by the same ratio.

4. How a Microphone Tuner Finds the Pitch

There are two common approaches to extracting pitch from a microphone signal. This tool uses autocorrelation.

The idea is more intuitive than the name suggests. Copy the waveform, slide the copy along, and check how well it overlaps the original. Overlap peaks when you've slid it by exactly one cycle. That slide distance is the period, and period gives you frequency.

frequency = sample rate ÷ period (in samples)

If you're recording at 48000Hz and one period is 582 samples, then 48000 ÷ 582 ≈ 82.4Hz — a guitar's low E2.

Parabolic interpolation is layered on top: the values on either side of the best-overlap point are used to estimate a fractional position between whole samples. Without it, periods land only on integers and readings jump in visible steps, especially up high.

5. 🔴 Why Low Notes Read an Octave Too High

Trying to tune a bass low E and getting E2 instead of E1 is a common experience. The app isn't broken — it's a limit of the analysis window.

Autocorrelation slides a copy of the waveform against itself, so it needs to be able to slide by at least one full period. The maximum slide is bounded by the window length. With the commonly used 2048-sample window searched to half its length, you get:

Sample rate Max lag Detection floor
44100Hz 1024 samples about 43.1Hz
48000Hz 1024 samples about 46.9Hz

A bass low E1 is 41.2Hz — below the floor in both cases.

When the true period falls outside the search range the signal doesn't vanish. The waveform still overlaps well at half that period, so that's what gets picked, and the reading lands an octave up at E2 (82.4Hz). It's the classic octave error — except under these conditions it isn't bad luck, it's structurally guaranteed.

Which is why it got fixed while writing this

This site's tuner was in exactly that state. Digging into the implementation for this article surfaced it, and it was fixed the same day.

  • The analysis window now scales with the instrument — lower presets get a longer window. It now resolves E1, and the 5-string bass B0 (30.87Hz) as well
  • Peak selection changed from "highest" to "first peak above a threshold." On harmonically rich notes the double period can score higher than the true one, which produced the opposite error — reading an octave low. That's covered now too
  • Unit tests on synthetic tones pin E1, B0, and harmonic-heavy signals as regression cases, so shortening the window again breaks a test first

If another tuner shows the same symptom

Plenty of free tuner apps still carry this limit. If low notes keep landing an octave high:

  • Use the 12th-fret harmonic — it rings an octave up, clearly inside any tuner's range
  • Tune at the 5th fret — fretting low E there gives A1 (55Hz), above most floors
  • The other strings (A1 55Hz · D2 73.4Hz · G2 98Hz) were never affected

Guitars are fine throughout: the lowest string is E2 at 82.41Hz, so all six sit well inside range. This only shows up on low bass notes.

6. Quiet Signals Are Ignored on Purpose

Sometimes the tuner shows nothing at all. That's deliberate, not broken.

If the incoming signal falls below a volume threshold, the calculation is skipped entirely. Without that guard, the algorithm would find some period in ambient noise and display a confident, meaningless note.

The catch is that silence tells you nothing — you can't tell "not hearing you" from "hearing you but can't lock on." So an input level meter went in alongside the fix above. If the bar moves but no note appears, the sound is getting through and the answer isn't to pluck harder — it's to let only one string ring.

Otherwise, to get a clean read:

  • Move closer to the mic and pluck a little harder
  • Wait about half a second before reading — the initial attack is a mess of overtones and the number wobbles
  • Work somewhere quiet — low-frequency hum like air conditioning interferes most

7. Recap

  • Every note is calculated from A4 = 440Hz, multiplying by 1.05946 per semitone
  • A cent is 1/100 of a semitone — one ruler that works in any register
  • ±5 cents for green is set near the limit of human discrimination. 440 vs 442 is 7.9 cents, which crosses it
  • 🔴 Low notes reading an octave high comes from the analysis window — a 2048-sample window puts the floor at 43–47Hz, which catches bass E1 (41.2Hz). This tool was fixed the same day it was found; elsewhere, use a harmonic or the 5th fret
  • Very quiet input is deliberately ignored. If the level meter moves but no note appears, the fix is letting one string ring — not plucking harder

A tuner's numbers aren't absolute truth — they show how far you are from an agreed reference. If the people you're playing with use a different one, matching them beats chasing the green light.

To see this yourself, open the tuner and move the reference from 440 to 442. Every target shifts without you touching an instrument.

If timing is the bigger problem, there's a metronome, and for ear training try the chromatic trainer.

Recommended Tools for You

🚀 JAY Project · 60+ Free Web Tools

60+ free web tools built as a hobby — no signup, no payment, every input stays in your browser.